U.S. patent application number 17/130676 was filed with the patent office on 2022-06-23 for dynamic inclusive last level cache.
This patent application is currently assigned to Intel Corporation. The applicant listed for this patent is Intel Corporation. Invention is credited to Ayan Mandal, Joseph Nuzman, Leon Polishuk, Oz Shitrit.
Application Number | 20220197797 17/130676 |
Document ID | / |
Family ID | |
Filed Date | 2022-06-23 |
United States Patent
Application |
20220197797 |
Kind Code |
A1 |
Mandal; Ayan ; et
al. |
June 23, 2022 |
DYNAMIC INCLUSIVE LAST LEVEL CACHE
Abstract
An embodiment of an integrated circuit may comprise a core, and
a cache controller coupled to the core, the cache controller
including circuitry to identify data from a working set for dynamic
inclusion in a next level cache based on an amount of re-use of the
next level cache, send a shared copy of the identified data to a
requesting core of one or more processor cores, and maintain a copy
of the identified data in the next level cache. Other embodiments
are disclosed and claimed.
Inventors: |
Mandal; Ayan; (Bangaluru,
IN) ; Polishuk; Leon; (Haifa, IL) ; Shitrit;
Oz; (Tel Aviv, IL) ; Nuzman; Joseph; (Haifa,
IL) |
|
Applicant: |
Name |
City |
State |
Country |
Type |
Intel Corporation |
Santa Clara |
CA |
US |
|
|
Assignee: |
Intel Corporation
Santa Clara
CA
|
Appl. No.: |
17/130676 |
Filed: |
December 22, 2020 |
International
Class: |
G06F 12/0811 20060101
G06F012/0811; G06F 12/128 20060101 G06F012/128; G06F 12/0815
20060101 G06F012/0815 |
Claims
1. An integrated circuit, comprising: a core; and a cache
controller coupled to the core, the cache controller including
circuitry to: identify data from a working set for dynamic
inclusion in a next level cache based on an amount of re-use of the
next level cache, send a shared copy of the identified data to a
requesting core of one or more processor cores, and maintain a copy
of the identified data in the next level cache.
2. The integrated circuit of claim 1, wherein the circuitry is
further to: determine dynamic inclusion of data in the next level
cache on a per data line basis.
3. The integrated circuit of claim 1, wherein the circuitry is
further to: increment a counter value when a hit in the next level
cache corresponds to an eviction from a core cache; and identify a
current data hit in the next level cache for dynamic inclusion in
the next level cache if the current data hit corresponds to an
eviction from the core cache and if the counter value is greater
than a threshold.
4. The integrated circuit of claim 3, wherein the circuitry is
further to: set a snoop filter to indicate that the requesting core
is valid for the current data hit.
5. The integrated circuit of claim 4, wherein, if the current data
hit does not correspond to an eviction from the core cache or if
the counter value is not greater than the threshold, the circuitry
is further to: send an exclusive copy of the data to the requesting
core; update an entry in the snoop filter to indicate a core
identifier of the requesting core; and deallocate the data in the
next level cache.
6. The integrated circuit of claim 1, wherein the circuitry is
further to: silently drop data to be evicted from a core cache if
the data to be evicted from the core cache has a shared copy of the
data in the next level cache.
7. The integrated circuit of claim 1, wherein the next level cache
comprises a non-inclusive last level cache.
8. A method of controlling a cache, comprising: identifying data
from a working set for dynamic inclusion in a next level cache
based on an amount of re-use of the next level cache; sending a
shared copy of the identified data to a requesting core of one or
more processor cores; and maintaining a copy of the identified data
in the next level cache.
9. The method of claim 8, further comprising: determining dynamic
inclusion of data in the next level cache on a per data line
basis.
10. The method of claim 8, further comprising: incrementing a
counter value when a hit in the next level cache corresponds to an
eviction from a core cache; and identifying a current data hit in
the next level cache for dynamic inclusion in the next level cache
if the current data hit corresponds to an eviction from the core
cache and if the counter value is greater than a threshold.
11. The method of claim 10, further comprising: setting a snoop
filter to indicate that the requesting core is valid for the
current data hit.
12. The method of claim 11, wherein, if the current data hit does
not correspond to an eviction from the core cache or if the counter
value is not greater than the threshold, the method further
comprises: sending an exclusive copy of the data to the requesting
core; updating an entry in the snoop filter to indicate a core
identifier of the requesting core; and deallocating the data in the
next level cache.
13. The method of claim 8, further comprising: silently dropping
data to be evicted from a core cache if the data to be evicted from
the core cache has a shared copy of the data in the next level
cache.
14. An apparatus, comprising: one or more processor cores; a core
cache co-located with and communicatively coupled to the one or
more processor cores; a next level cache co-located with and
communicatively coupled to the core cache and the one or more
processor cores; and a cache controller co-located with and
communicatively coupled to the core cache, the next level cache,
and the one or more processor cores, the cache controller including
circuitry to: identify data from a working set for dynamic
inclusion in the next level cache based on an amount of re-use of
the next level cache, send a shared copy of the identified data to
a requesting core of the one or more processor cores, and maintain
a copy of the identified data in the next level cache.
15. The apparatus of claim 14, wherein the circuitry is further to:
determine dynamic inclusion of data in the next level cache on a
per data line basis.
16. The apparatus of claim 14, wherein the circuitry is further to:
increment a counter value when a hit in the next level cache
corresponds to an eviction from the core cache; and identify a
current data hit in the next level cache for dynamic inclusion in
the next level cache if the current data hit corresponds to an
eviction from the core cache and if the counter value is greater
than a threshold.
17. The apparatus of claim 16, wherein the circuitry is further to:
set a snoop filter to indicate that the requesting core is valid
for the current data hit.
18. The apparatus of claim 16, wherein, if the current data hit
does not correspond to an eviction from the core cache or if the
counter value is not greater than the threshold, the circuitry is
further to: send an exclusive copy of the data to the requesting
core; update an entry in the snoop filter to indicate a core
identifier of the requesting core; and deallocate the data in the
next level cache.
19. The apparatus of claim 14, wherein the circuitry is further to:
silently drop data to be evicted from a core cache if the data to
be evicted from the core cache has a shared copy of the data in the
next level cache.
20. The apparatus of claim 14, wherein the next level cache
comprises a non-inclusive last level cache.
Description
BACKGROUND
1. Technical Field
[0001] This disclosure generally relates to processor technology,
and processor cache technology.
2. Background Art
[0002] For an integrated circuit chip/package that includes a
processor, a last level cache (LLC) may refer to a highest-level
cache that may be shared by all the functional units in the same
chip/package with the LLC. LLC cache can be classified based on
whether the inclusion policy is inclusive, exclusive, or
non-inclusive. If all the blocks that are present in the core
caches (e.g., mid-level cache (MLC) and first-level (L1) cache) are
also present in the LLC, then the LLC is considered inclusive of
the core caches. If the LLC only contains blocks that are not
present in the core caches, then the LLC is considered exclusive of
the core caches. An exclusive LLC policy reduces memory accesses by
effectively utilizing a combined capacity of the core caches and
the LLC, as compared to an inclusive LLC policy where the capacity
of the LLC determines the overall capacity because the blocks are
duplicated between the core caches and the LLC.
[0003] Exclusive LLC may require additional on-chip bandwidth to
support more frequent evictions (e.g., clean as well as modified)
from the core caches. For inclusive LLC, the core caches may
silently drop a clean eviction from the core caches because a copy
of the evicted line already exists in the LLC. A non-inclusive LLC
policy (sometimes also referred to as non-inclusive non-exclusive
(NINE)) does not enforce either inclusion or exclusion. For
example, the LLC may contain blocks from the core caches but the
non-inclusive LLC policy does not provide any guarantee on the data
duplication between the two.
BRIEF DESCRIPTION OF THE DRAWINGS
[0004] The various embodiments of the present invention are
illustrated by way of example, and not by way of limitation, in the
figures of the accompanying drawings and in which:
[0005] FIG. 1 is a block diagram of an example of an integrated
circuit according to an embodiment;
[0006] FIGS. 2A to 2C are flow diagrams of an example of a method
of controlling a cache according to an embodiment;
[0007] FIG. 3 is a block diagram of an example of an apparatus
according to an embodiment;
[0008] FIG. 4 is a flow diagram of an example of a process flow
according to an embodiment;
[0009] FIG. 5 is a flow diagram of another example of a process
flow according to an embodiment;
[0010] FIG. 6 is a block diagram of an example of a cache system
according to an embodiment;
[0011] FIG. 7 is a block diagram of another example of an
integrated circuit according to an embodiment;
[0012] FIGS. 8A to 8C are flow diagrams of another example of a
method of controlling a cache according to an embodiment;
[0013] FIG. 9 is a block diagram of an example of another apparatus
according to an embodiment;
[0014] FIG. 10 is an illustrative diagram of example of memory
access patterns according to an embodiment;
[0015] FIG. 11 is a flow diagram of another example of a method of
controlling a cache according to an embodiment;
[0016] FIG. 12 is a block diagram of another example of a cache
system according to an embodiment;
[0017] FIG. 13A is a block diagram illustrating both an exemplary
in-order pipeline and an exemplary register renaming, out-of-order
issue/execution pipeline according to embodiments of the
invention.
[0018] FIG. 13B is a block diagram illustrating both an exemplary
embodiment of an in-order architecture core and an exemplary
register renaming, out-of-order issue/execution architecture core
to be included in a processor according to embodiments of the
invention;
[0019] FIGS. 14A-B illustrate a block diagram of a more specific
exemplary in-order core architecture, which core would be one of
several logic blocks (including other cores of the same type and/or
different types) in a chip;
[0020] FIG. 15 is a block diagram of a processor that may have more
than one core, may have an integrated memory controller, and may
have integrated graphics according to embodiments of the
invention;
[0021] FIGS. 16-19 are block diagrams of exemplary computer
architectures; and
[0022] FIG. 20 is a block diagram contrasting the use of a software
instruction converter to convert binary instructions in a source
instruction set to binary instructions in a target instruction set
according to embodiments of the invention.
DETAILED DESCRIPTION
[0023] Embodiments discussed herein variously provide techniques
and mechanisms for controlling a processor cache. The technologies
described herein may be implemented in one or more electronic
devices. Non-limiting examples of electronic devices that may
utilize the technologies described herein include any kind of
mobile device and/or stationary device, such as cameras, cell
phones, computer terminals, desktop computers, electronic readers,
facsimile machines, kiosks, laptop computers, netbook computers,
notebook computers, internet devices, payment terminals, personal
digital assistants, media players and/or recorders, servers (e.g.,
blade server, rack mount server, combinations thereof, etc.),
set-top boxes, smart phones, tablet personal computers,
ultra-mobile personal computers, wired telephones, combinations
thereof, and the like. More generally, the technologies described
herein may be employed in any of a variety of electronic devices
including integrated circuitry which is operable to control or
utilize a processor cach.
[0024] In the following description, numerous details are discussed
to provide a more thorough explanation of the embodiments of the
present disclosure. It will be apparent to one skilled in the art,
however, that embodiments of the present disclosure may be
practiced without these specific details. In other instances,
well-known structures and devices are shown in block diagram form,
rather than in detail, in order to avoid obscuring embodiments of
the present disclosure.
[0025] Note that in the corresponding drawings of the embodiments,
signals are represented with lines. Some lines may be thicker, to
indicate a greater number of constituent signal paths, and/or have
arrows at one or more ends, to indicate a direction of information
flow. Such indications are not intended to be limiting. Rather, the
lines are used in connection with one or more exemplary embodiments
to facilitate easier understanding of a circuit or a logical unit.
Any represented signal, as dictated by design needs or preferences,
may actually comprise one or more signals that may travel in either
direction and may be implemented with any suitable type of signal
scheme.
[0026] Throughout the specification, and in the claims, the term
"connected" means a direct connection, such as electrical,
mechanical, or magnetic connection between the things that are
connected, without any intermediary devices. The term "coupled"
means a direct or indirect connection, such as a direct electrical,
mechanical, or magnetic connection between the things that are
connected or an indirect connection, through one or more passive or
active intermediary devices. The term "circuit" or "module" may
refer to one or more passive and/or active components that are
arranged to cooperate with one another to provide a desired
function. The term "signal" may refer to at least one current
signal, voltage signal, magnetic signal, or data/clock signal. The
meaning of "a," "an," and "the" include plural references. The
meaning of "in" includes "in" and "on."
[0027] The term "device" may generally refer to an apparatus
according to the context of the usage of that term. For example, a
device may refer to a stack of layers or structures, a single
structure or layer, a connection of various structures having
active and/or passive elements, etc. Generally, a device is a
three-dimensional structure with a plane along the x-y direction
and a height along the z direction of an x-y-z Cartesian coordinate
system. The plane of the device may also be the plane of an
apparatus which comprises the device.
[0028] The term "scaling" generally refers to converting a design
(schematic and layout) from one process technology to another
process technology and subsequently being reduced in layout area.
The term "scaling" generally also refers to downsizing layout and
devices within the same technology node. The term "scaling" may
also refer to adjusting (e.g., slowing down or speeding up--i.e.
scaling down, or scaling up respectively) of a signal frequency
relative to another parameter, for example, power supply level.
[0029] The terms "substantially," "close," "approximately," "near,"
and "about," generally refer to being within +/-10% of a target
value. For example, unless otherwise specified in the explicit
context of their use, the terms "substantially equal," "about
equal" and "approximately equal" mean that there is no more than
incidental variation between among things so described. In the art,
such variation is typically no more than +/-10% of a predetermined
target value.
[0030] It is to be understood that the terms so used are
interchangeable under appropriate circumstances such that the
embodiments of the invention described herein are, for example,
capable of operation in other orientations than those illustrated
or otherwise described herein.
[0031] Unless otherwise specified the use of the ordinal adjectives
"first," "second," and "third," etc., to describe a common object,
merely indicate that different instances of like objects are being
referred to and are not intended to imply that the objects so
described must be in a given sequence, either temporally,
spatially, in ranking or in any other manner.
[0032] The terms "left," "right," "front," "back," "top," "bottom,"
"over," "under," and the like in the description and in the claims,
if any, are used for descriptive purposes and not necessarily for
describing permanent relative positions. For example, the terms
"over," "under," "front side," "back side," "top," "bottom,"
"over," "under," and "on" as used herein refer to a relative
position of one component, structure, or material with respect to
other referenced components, structures or materials within a
device, where such physical relationships are noteworthy. These
terms are employed herein for descriptive purposes only and
predominantly within the context of a device z-axis and therefore
may be relative to an orientation of a device. Hence, a first
material "over" a second material in the context of a figure
provided herein may also be "under" the second material if the
device is oriented upside-down relative to the context of the
figure provided. In the context of materials, one material disposed
over or under another may be directly in contact or may have one or
more intervening materials. Moreover, one material disposed between
two materials may be directly in contact with the two layers or may
have one or more intervening layers. In contrast, a first material
"on" a second material is in direct contact with that second
material. Similar distinctions are to be made in the context of
component assemblies.
[0033] The term "between" may be employed in the context of the
z-axis, x-axis or y-axis of a device. A material that is between
two other materials may be in contact with one or both of those
materials, or it may be separated from both of the other two
materials by one or more intervening materials. A material
"between" two other materials may therefore be in contact with
either of the other two materials, or it may be coupled to the
other two materials through an intervening material. A device that
is between two other devices may be directly connected to one or
both of those devices, or it may be separated from both of the
other two devices by one or more intervening devices.
[0034] As used throughout this description, and in the claims, a
list of items joined by the term "at least one of" or "one or more
of" can mean any combination of the listed terms. For example, the
phrase "at least one of A, B or C" can mean A; B; C; A and B; A and
C; B and C; or A, B and C. It is pointed out that those elements of
a figure having the same reference numbers (or names) as the
elements of any other figure can operate or function in any manner
similar to that described, but are not limited to such.
[0035] In addition, the various elements of combinatorial logic and
sequential logic discussed in the present disclosure may pertain
both to physical structures (such as AND gates, OR gates, or XOR
gates), or to synthesized or otherwise optimized collections of
devices implementing the logical structures that are Boolean
equivalents of the logic under discussion.
[0036] Dynamic Inclusion Cache Policy Examples
[0037] Some embodiments provide technology for a dynamic inclusive
LLC (DIL). As noted above, exclusive LLC provides additional
capacity over inclusive LLC at the cost of additional data transfer
from the MLC to the LLC for the MLC clean evictions and additional
power consumption. Workloads with a majority of the working set
(e.g., amount of data accessed in a given time window) that fits
within the capacity of the LLC show higher power consumption under
the exclusive LLC policy as compared to the inclusive LLC policy.
In order to ensure that the LLC stays as a point of coherence, each
time the MLC sends an eviction, the MLC needs to enquire the first
level cache (L1) to find out if the line exists in the first level
cache. Such backward enquiry from the MLC to the first level cache
for every MLC eviction adds an additional pressure on the MLC
controller bandwidth. Workloads that may already be bottlenecked by
the LLC read bandwidth get an additional bottleneck from the MLC
clean eviction bandwidth, as compared to inclusive LLC. Using
inclusive LLC always retains a copy of the line in LLC and notifies
the core to drop the clean evictions. Inclusive LLC, however,
removes the additive capacity of the LLC and the MLC from the
exclusive LLC policy and may cause performance loss for workloads
sensitive to the additional combined capacity of the MLC and the
LLC.
[0038] Some embodiments may advantageously provide technology for
dynamic inclusivity for LLC to gain benefits of both the inclusive
LLC (e.g., low data transfer between the MLC and the LLC and no
penalty of doing L1 backward enquire on each MLC eviction) and the
exclusive LLC (e.g., combined capacity of the MLC and the LLC).
Some embodiments of DIL may provide technology to identify
workloads that get a high re-use from the LLC, to send a shared
copy of the data to the core and, at the same time, maintain a copy
of the data in the LLC. When the MLC needs to evict the data from
the cache, the MLC may silently drop the data to be evicted because
the data has a shared copy and the LLC already holds the data. The
shared copy maintained in the LLC avoids the additional data
transfer from the MLC to the LLC and thereby saves power. The
shared copy maintained in the LLC also saves the effort of back
invalidating L1 for every MLC clean eviction and improves the
second level cache (L2) throughput significantly for workloads
showing significant re-use from the LLC.
[0039] Advantageously, embodiments of DIL may improve a LLC peak
bandwidth significantly by reducing or eliminating the need of
snooping L1 for each MLC eviction. Embodiments of DIL may also
improve the LLC power and according the package power
significantly, which may lead to processors with better performance
and throughput characteristics.
[0040] With reference to FIG. 1, an embodiment of an integrated
circuit 100 may include a core 111, and a cache controller 112
coupled to the core 111. The cache controller 112 may include
circuitry 113 to identify data from a working set for dynamic
inclusion in a next level cache 114 based on an amount of re-use of
the next level cache 114, send a shared copy of the identified data
to a requesting core of one or more processor cores, and maintain a
copy of the identified data in the next level cache 114. For
example, the circuitry 113 may be configured to determine dynamic
inclusion of data in the next level cache 114 on a per data line
basis. In some embodiments, the circuitry 113 may be further
configured to silently drop data to be evicted from a core cache
115 if the data to be evicted from the core cache 115 has a shared
copy of the data in the next level cache 114. For example, the next
level cache 114 may comprise a non-inclusive LLC.
[0041] In some embodiments, the circuitry 113 may be further
configured to increment a counter value when a hit in the next
level cache 114 corresponds to an eviction from a core cache 115,
and identify a current data hit in the next level cache 114 for
dynamic inclusion in the next level cache 114 if the current data
hit corresponds to an eviction from the core cache 115 and if the
counter value is greater than a threshold. For example, the
circuitry 113 may also be configured to set a snoop filter to
indicate that the requesting core is valid for the current data
hit. In some embodiments, if the current data hit does not
correspond to an eviction from the core cache or if the counter
value is not greater than the threshold, the circuitry 113 may be
further configured to send an exclusive copy of the data to the
requesting core, update an entry in the snoop filter to indicate a
core identifier of the requesting core, and deallocate the data in
the next level cache 114.
[0042] Embodiments of the cache controller 112, circuitry 113, next
level cache 114, and/or core cache 115 may be incorporated in a
processor including, for example, the core 990 (FIG. 13B), the
cores 1102A-N (FIGS. 15, 19), the processor 1210 (FIG. 16), the
co-processor 1245 (FIG. 16), the processor 1370 (FIGS. 17-18), the
processor/coprocessor 1380 (FIGS. 17-18), the coprocessor 1338
(FIGS. 17-18), the coprocessor 1520 (FIG. 19), and/or the
processors 1614, 1616 (FIG. 20).
[0043] With reference to FIGS. 2A to 2C, an embodiment of a method
200 of controlling a cache may include identifying data from a
working set for dynamic inclusion in a next level cache based on an
amount of re-use of the next level cache at box 211, sending a
shared copy of the identified data to a requesting core of one or
more processor cores at box 212, and maintaining a copy of the
identified data in the next level cache at box 213. For example,
the method 200 may include determining dynamic inclusion of data in
the next level cache on a per data line basis at box 214. Some
embodiments of the method 200 may further include silently dropping
data to be evicted from a core cache if the data to be evicted from
the core cache has a shared copy of the data in the next level
cache at box 215. For example, the next level cache may comprise a
non-inclusive LLC at box 216.
[0044] Some embodiments of the method 200 may further include
incrementing a counter value when a hit in the next level cache
corresponds to an eviction from a core cache at box 217, and, if a
current data hit corresponds to an eviction from the core cache and
if the counter value is greater than a threshold at box 218,
identifying the current data hit in the next level cache for
dynamic inclusion in the next level cache at box 219. The method
200 may also include setting a snoop filter to indicate that the
requesting core is valid for the current data hit at box 220. In
some embodiments, if the current data hit does not correspond to an
eviction from the core cache or if the counter value is not greater
than the threshold at box 218, the method 200 may further include
sending an exclusive copy of the data to the requesting core at box
221, updating an entry in the snoop filter to indicate a core
identifier of the requesting core at box 222, and deallocating the
data in the next level cache at box 223.
[0045] With reference to FIG. 3, an embodiment of an apparatus 300
may include one or more processor cores 332, a core cache 333
co-located with and communicatively coupled to the one or more
processor cores 332, a next level cache 334 co-located with and
communicatively coupled to the core cache 333 and the one or more
processor cores 332, and a cache controller 335 co-located with and
communicatively coupled to the core cache 333, the next level cache
334, and the one or more processor cores 332. Any suitable
technology may be utilized for the connections between the
components of the apparatus 300 including, for example, bus, ring,
other fabric, etc. The cache controller 335 may include DIL
circuitry 336. The circuitry 336 may be configured to identify data
from a working set for dynamic inclusion in the next level cache
334 based on an amount of re-use of the next level cache 334, send
a shared copy of the identified data to a requesting core of the
one or more processor cores 332, and maintain a copy of the
identified data in the next level cache 334. For example, the
circuitry 336 may be configured to determine dynamic inclusion of
data in the next level cache 334 on a per data line basis. In some
embodiments, the circuitry 336 may be further configured to
silently drop data to be evicted from the core cache 333 if the
data to be evicted from the core cache 333 has a shared copy of the
data in the next level cache 334. For example, the next level cache
334 may comprise a non-inclusive LLC.
[0046] In some embodiments of the apparatus 300, the circuitry 336
may be further configured to increment a counter value when a hit
in the next level cache 334 corresponds to an eviction from the
core cache 333, and identify a current data hit in the next level
cache 334 for dynamic inclusion in the next level cache 334 if the
current data hit corresponds to an eviction from the core cache 333
and if the counter value is greater than a threshold. The circuitry
336 may also be configured to set a snoop filter to indicate that
the requesting core is valid for the current data hit. In some
embodiments, if the current data hit does not correspond to an
eviction from the core cache 333 or if the counter value is not
greater than the threshold, the circuitry 336 may be further
configured to send an exclusive copy of the data to the requesting
core, update an entry in the snoop filter to indicate a core
identifier of the requesting core, and deallocate the data in the
next level cache 334.
[0047] Embodiments of the cache controller 335, DIL circuitry 336,
next level cache 334, and/or core cache 333 may be integrated with
processors including, for example, the core 990 (FIG. 13B), the
cores 1102A-N (FIGS. 15, 19), the processor 1210 (FIG. 16), the
co-processor 1245 (FIG. 16), the processor 1370 (FIGS. 17-18), the
processor/coprocessor 1380 (FIGS. 17-18), the coprocessor 1338
(FIGS. 17-18), the coprocessor 1520 (FIG. 19), and/or the
processors 1614, 1616 (FIG. 20).
[0048] As noted above, for an exclusive LLC, each MLC clean
eviction needs to send the data to the LLC because the block was
present only in the MLC. This additional data transfer causes
additional power consumption in the chip/package (e.g., an SoC
package) as compared to the inclusive LLC. The non-inclusive LLC on
the other hand, provides no guarantees on the data duplication
between the core caches and the LLC. A non-inclusive LLC may be
configured to insert blocks into either the MLC, or the LLC, or
both. A conventional non-inclusive LLC may provide the following
process flows: A) for a read LLC miss, the data is installed only
in the MLC; B) for a read LLC hit, the line is deallocated from the
LLC and allocated in the MLC; and C) the MLC sends both clean and
modified evictions to LLC.
[0049] A non-inclusive LLC may also include a snoop filter (SF)
which behaves as an inclusive LLC but without any data storage. The
SF enables the LLC to provide coherence without additional snoop
overhead. In some conventional non-inclusive LLCs, for example, any
miss in the LLC does not guarantee that any core does not have the
line and the cache controller need a snoop to all the cores. The SF
avoids these broadcast snoops by maintaining the tags of all the
lines that are present in all the cores. Because the SF does not
have any data storage, the SF may be a light weight circuit in
terms of the area and power consumption. Some processor
chips/packages may utilize a common tag storage for both the SF and
the LLC data. For example, each tag entry may contain the following
major information: a) a core valid field (e.g., that indicates
which core caches may have the line); b) a data valid field (e.g.,
that indicates if the LLC contains the data); and c) a state field
(e.g., that indicates a state of the cache line either in MLC or
LLC with respect to DRAM).
[0050] One example reason that a core demand read request in the
non-inclusive LLC hits in the LLC is because the data line was
first issued as an LLC pre-fetch and later the core demand read got
a hit to the pre-fetched data in the LLC. In this scenario, the LLC
acts as a pre-fetch buffer and hides the memory latency but does
not save on the memory access for the given data line. Another
example reason that a core demand read request in the non-inclusive
LLC hits in the LLC is when the core demand read request gets a hit
to a previous MLC eviction from either the same core or a different
core and the LLC acts as a victim cache. For this scenario, the LLC
provides the re-use of the data line and accordingly saves the
memory access. The cache controller may maintain a counter referred
to as the LLC hit counter (LHC) which captures this re-use from the
LLC and is incremented on every LLC hit to an earlier MLC eviction
in the LLC. A high value of the LHC may indicate that the working
set of the application fits within the LLC. Accordingly, a high
value of LHC may indicate that an inclusive LLC might perform
better for that working set because the inclusive LLC may provide
at least the following benefits: a) the MLC need not snoop L1 on
every clean eviction, improving MLC controller bandwidth; and b)
the clean eviction is dropped from the MLC, saving on the write
bandwidth from the MLC to the LLC.
[0051] With reference to FIG. 4, an embodiment of a process flow
400 shows an example of a core demand read request flow in a
non-inclusive LLC with DIL. At box 411, a LLC lookup determines if
the line is present in the LLC. Conventionally, if the line is
present in the LLC, an exclusive copy of the data is sent to the
requesting core, the SF entry is updated with the core id of the
requesting core in the core valid field, and the data entry is
deallocated. In accordance with some embodiment of DIL, when there
is a LLC hit at box 412, the cache controller may then determine
whether the data was brought into the LLC by an earlier MLC
Eviction and whether the LHC value is greater than a threshold at
box 413. When both the above conditions are met at box 413,
indicating a high re-use probability from the LLC of the given
line, the cache controller then sends a shared copy of the line to
the core at box 414 and the cache controller keeps the LLC data
along with the SF entry at box 415 (e.g., the LLC data entry is not
deallocated). The SF may then be populated with the core valid bit
of the requesting core. If the two conditions are not met at box
413, the cache controller may proceed to send an exclusive copy of
the data to the core at box 416, and evict the LLC data and keep
only the SF entry at box 417.
[0052] With reference to FIG. 5, an embodiment of a process flow
440 shows an example of a MLC Eviction flow for a clean victim. For
a conventional non-inclusive LLC, when the victim is exclusive, the
core needs to send the copy of the data back to the LLC. In the
conventional process, the core valid entry of the SF entry is
cleared as well. Accordingly, the MLC eviction must snoop the L1 to
check if the line is present in the L1. When the line is not
present in L1 (common case), the clean eviction is sent back to the
LLC which populates the LLC data entry and clears the core valid
bit. In a conventional corner case when the line is present in L1,
only the data entry in LLC is populated but the core valid bit is
not cleared.
[0053] In the process flow 400, however, the LLC sends a shared
copy of the line to the MLC when the application working set fits
in LLC. At MLC eviction at box 441, for a clean victim at box 442,
the cache controller may determine if the state of the clean victim
is shared at box 443. If so, because the clean victim is a shared
copy and the LLC already has a copy of the data, the MLC drops the
data silently at box 444. Dropping the data silently saves both in
the data transfer from MLC and LLC and also saves the L1 snoop, as
compared to the conventional MLC eviction for a non-inclusive LLC.
If the state is not shared at box 443, the cache controller may
proceed to snoop L1 at box 445 and evict the data to LLC at box
446.
[0054] In terms of design complexity, there is no change needed in
the core because the inclusivity is facilitated by sending a shared
copy of the line to the core and the information of whether LLC is
behaving as inclusive or exclusive is not propagated to the MLC.
Embodiments also advantageously avoid any transition overhead
between inclusive and exclusive behavior. In some embodiments, the
cache controller for the LLC may determine the inclusivity per data
line and accordingly there is no need for synchronization across
MLC and LLC. If the core needs to modify the data, however, the
core needs an exclusive copy, which incurs an additional request
from MLC to the LLC. Performance modeling of an embodiment of DIL
on a variety of standard micro-benchmarks which measures LLC
bandwidth showed better LLC peak bandwidth for different read-write
mixes on a single core and multi-core applications, and better
write bandwidth and instructions per cycle (IPC) for multi-thread
applications, versus a baseline non-inclusive LLC without DIL
[0055] With reference to FIG. 6, an embodiment of a cache system
460 includes a cache controller 462 communicatively coupled to a
core cache 464, a LLC 466, and a SF 468. The core cache 464
includes a L1 cache 464a and a MLC 464b. The cache controller 462
maintains a LHC which is incremented on every LLC hit to an earlier
MLC eviction in the LLC 466. The cache controller 462 is configured
with DIL technology to handle a core demand read for a data line as
follows. The cache controller 462 performs a LLC lookup to
determine if the data line is present in the LLC 466. If there is a
hit in the LLC 466, the cache controller 462 then determines
whether the data was brought into the LLC 466 by an earlier
eviction from the MLC 464b and whether the LHC value is greater
than a threshold, indicating a high re-use probability of the data
line from the LLC 466. If both conditions are met, the cache
controller 462 then sends a shared copy of the data line to the
core, the cache controller 462 keeps the data line in the LLC 466,
and also keeps the corresponding entry in the SF 468 (e.g., the LLC
data entry is not deallocated). The SF 468 is then populated with
the core valid bit of the requesting core. If the two conditions
are not met, the cache controller 462 proceeds to send an exclusive
copy of the data to the core, evict the data line from the LLC 466,
and keep only the corresponding entry in the SF 468.
[0056] Some embodiments of the cache controller 462 are further
configured with DIL technology to handle an eviction from the MLC
464b for a clean victim as follows. The cache controller determines
if the state of the clean victim is shared and, if so, the cache
controller silently drops the data from the MLC 464b (e.g., because
the clean victim is a shared copy and the LLC 466 already has a
copy of the data). Dropping the data silently saves both in the
data transfer from the MLC 464b and the LLC 466 and also saves a
snoop of the L1 cache 464a, as compared to the conventional MLC
eviction for a non-inclusive LLC. If the state of the clean victim
is not shared, the cache controller 462 proceed to snoop the L1
cache 464a (e.g., updating the corresponding entry in the SF 468)
and evict the data to the LLC 466.
[0057] Single Re-Use Cache Policy Examples
[0058] Some embodiments provide technology to apply or enforce a
single re-use cache policy. For exclusive LLC, the LLC may be used
as a victim cache where all the MLC evictions are copied back to
the LLC with the expectation of getting re-used from the LLC in the
future. However, not all the MLC evictions have equal probability
of getting re-used from the LLC. Some systems may utilize dead
block prediction (DBP) techniques to bypass some of the MLC
evictions to prevent LLC thrashing and provide improved or optimal
LLC re-use. Conventional DBP techniques for exclusive LLC, however,
may not effectively capture single re-use data from the LLC (e.g.,
data read from the main memory for the first time and then re-used
a second time), which may result in a lower LLC hit rate and lower
performance. For example, an exclusive LLC with DBP may not capture
the single re-use of a buffer even if the buffer capacity is
smaller than the LLC size.
[0059] Some embodiments may provide technology for a single re-use
policy (SRP), where a specific class of MLC evictions (e.g., with
the source as the main memory) may be given a second chance to stay
in LLC based on overall LLC re-use. Advantageously, some
embodiments of SRP technology may significantly improve the LLC hit
rate of certain applications, thereby reducing the main memory
access.
[0060] With reference to FIG. 7, an embodiment of an integrated
circuit 500 may include a core 511, and a cache controller 512
coupled to the core 511. The cache controller 512 may include
circuitry 513 to identify single re-use data evicted from a core
cache 514, and retain the identified single re-use data in a next
level cache 515 based on an overall re-use of the next level cache
515. For example, a source of the single re-use data may be main
memory. In some embodiments, the circuitry 513 may be configured to
determine a use count for a data line based on a number of core
cache 514 hits experienced by the data line when the data line is
resident in the core cache 514, determine a trip count for the data
line based on a number of trips made by the data line between the
core cache 514 and the next level cache 515 from when the data line
is brought into one or more of the core cache 514 and the next
level cache 515 until the data line is evicted from the next level
cache 515, and identify the single re-use data based on a use count
of one and trip count of zero.
[0061] In some embodiments, the circuitry 513 may be further
configured to increment a counter value when a hit in the next
level cache corresponds to an eviction from the core cache. The
circuitry 513 may also be configured to evict a data line from the
core cache 514, mark the evicted data line as dead, and install the
evicted data line marked as dead as a most recently used (MRU) data
line in the next level cache 515 if the counter value is greater
than a threshold and if a source of the data line is main memory.
In some embodiments, if the counter value is not greater than the
threshold or if a source of the data line is not main memory, the
circuitry 513 may be configured to install the evicted data line
marked as dead as a least recently used (LRU) data line in the next
level cache 515, if an invalid block is available in the next level
cache 515, or to bypass the next level cache 515, if an invalid
block is not available in the next level cache 515. For example,
the next level cache 515 may comprise a LLC.
[0062] Embodiments of the cache controller 512, circuitry 513, next
level cache 515, and/or core cache 514 may be incorporated in a
processor including, for example, the core 990 (FIG. 13B), the
cores 1102A-N (FIGS. 15, 19), the processor 1210 (FIG. 16), the
co-processor 1245 (FIG. 16), the processor 1370 (FIGS. 17-18), the
processor/coprocessor 1380 (FIGS. 17-18), the coprocessor 1338
(FIGS. 17-18), the coprocessor 1520 (FIG. 19), and/or the
processors 1614, 1616 (FIG. 20).
[0063] With reference to FIGS. 8A to 8C, an embodiment of a method
520 of controlling a cache may include identifying single re-use
data evicted from a core cache at box 521, and retaining the
identified single re-use data in a next level cache based on an
overall re-use of the next level cache at box 522. For example, a
source of the single re-use data may be main memory at box 523.
Some embodiments of the method 520 may further include determining
a use count for a data line based on a number of core cache hits
experienced by the data line when the data line is resident in the
core cache at box 524, determining a trip count for the data line
based on a number of trips made by the data line between the core
cache and the next level cache from when the data line is brought
into one or more of the core cache and the next level cache until
the data line is evicted from the next level cache at box 525, and
identifying the data line as single re-use data based on a use
count of one and trip count of zero at box 526.
[0064] Some embodiments of the method 520 may further include
incrementing a counter value when a hit in the next level cache
corresponds to an eviction from the core cache at box 527. The
method 520 may also include evicting a data line from the core
cache at box 528, marking the evicted data line as dead at box 529,
and, if the counter value is greater than a threshold and if a
source of the data line is main memory at box 530, installing the
evicted data line marked as dead as a MRU data line in the next
level cache at box 531. In some embodiments, if the counter value
is not greater than the threshold or if a source of the data line
is not main memory at box 530, the method 520 may further include
installing the evicted data line marked as dead as a LRU data line
in the next level cache at box 532 if an invalid block is available
in the next level cache, or bypassing the next level cache at box
533 if an invalid block is not available in the next level cache.
For example, the next level cache may comprise a LLC at box
534.
[0065] With reference to FIG. 9, an embodiment of an apparatus 540
may include one or more processor cores 542, a core cache 543
co-located with and communicatively coupled to the one or more
processor cores 542, a next level cache 544 co-located with and
communicatively coupled to the core cache 543 and the one or more
processor cores 542, and a cache controller co-located with and
communicatively coupled to the core cache 543, the next level cache
544, and the one or more processor cores 542. The cache controller
545 may include SRP circuitry 546. The circuitry 546 may be
configured to identify single re-use data evicted from the core
cache 543, and retain the identified single re-use data in the next
level cache 544 based on an overall re-use of the next level cache
544. For example, a source of the single re-use data is main
memory. In some embodiments, the circuitry 546 may be further
configured to determine a use count for a data line based on a
number of core cache 543 hits experienced by the data line when the
data line is resident in the core cache 543, determine a trip count
for the data line based on a number of trips made by the data line
between the core cache 543 and the next level cache 544 from when
the data line is brought into one or more of the core cache 543 and
the next level cache 544 until the data line is evicted from the
next level cache 544, and identify the single re-use data based on
a use count of one and trip count of zero.
[0066] In some embodiments, the circuitry 546 may be configured to
increment a counter value when a hit in the next level cache 544
corresponds to an eviction from the core cache 543. The circuitry
546 may also be configured to evict a data line from the core cache
543, mark the evicted data line as dead, and install the evicted
data line marked as dead as a most recently used data line in the
next level cache 544 if the counter value is greater than a
threshold and if a source of the data line is main memory. In some
embodiments, if the counter value is not greater than the threshold
or if a source of the data line is not main memory, the circuitry
546 may be further configured to install the evicted data line
marked as dead as a least recently used data line in the next level
cache if an invalid block is available in the next level cache, and
to bypass the next level cache if an invalid block is not available
in the next level cache. For example, the next level cache 544 may
comprise a LLC.
[0067] Embodiments of the cache controller 545, SRP circuitry 546,
next level cache 544, and/or core cache 543 may be integrated with
processors including, for example, the core 990 (FIG. 13B), the
cores 1102A-N (FIGS. 15, 19), the processor 1210 (FIG. 16), the
co-processor 1245 (FIG. 16), the processor 1370 (FIGS. 17-18), the
processor/coprocessor 1380 (FIGS. 17-18), the coprocessor 1338
(FIGS. 17-18), the coprocessor 1520 (FIG. 19), and/or the
processors 1614, 1616 (FIG. 20).
[0068] With reference to FIG. 10, an example diagram illustrates
different types of memory access patterns that an application may
shows and how LLC provides re-use for each of them. In this
example, the MLC capacity is 1.25 MB and the LLC capacity is 12 MB.
For a "Streaming" scenario, the core reads a new buffer (D, C, B,
A) of different capacity every time from the main memory. All these
accesses will be cold misses and the hit rate in the LLC will be
zero for any LLC size. Next, FIG. 10 shows a "Single Re-Use"
scenario where each buffer is read exactly twice from the main
memory. "D1" and "D2" represent two instances of the same buffer
"D" and are accessed by the core in the same order from the start
of the buffer "D" to the end of the buffer "D". Because the size of
the buffer "D" (20 MB) is bigger than the capacity of the LLC (12
MB) in this example, "D1" becomes a cold miss and "D2" becomes a
capacity miss. Similarly, "C1" and "C2" goes over the same buffer
"C" in the exact same order from the start of the buffer "C" to the
end of the buffer "C". Because the buffer capacity of "C" (2 MB) is
less than the LLC size (12 MB), the first iteration ("C1") is
expected to be a cold miss and the second iteration ("C2") is
expected to be an LLC Hit. Next, FIG. 10 shows a "Multi Re-Use"
scenario. Here the buffer "C" is accessed three times in terms of
"C1", "C2" and "C3" in the same order from the start of the buffer
"C" to the end of the buffer "C". C1 will be a cold miss and "C2"
and "C3" are expected to be LLC Hits because the capacity of "C" (2
MB) is less than the size of the LLC (12 MB).
[0069] The streaming scenarios issue unnecessary insertions (e.g.,
dead blocks) in the LLC that may waste on-chip bandwidth while not
improving performance. Any suitable technology may be utilized to
reduce the number of dead blocks in the LLCs. Example technology
may include techniques to improve a cache replacement algorithm,
techniques to bypass the LLC to save on-chip bandwidth, etc. Other
techniques may corelate the instruction or data addresses with the
death of a cache block (e.g., by utilizing the dead block either as
a replacement or as a prefetch target). Another technique may
utilize a virtual victim cache which uses the predicted dead blocks
to hold blocks evicted from other sets, where the second reference
to the evicted blocks may be satisfied from the dead pool instead
of going to the main memory.
[0070] Alternatively, other suitable technology to reduce the
number of dead blocks in the LLC may include techniques to use dead
block identification to bypass the LLC. The core tries to prevent
LLC thrashing by bypassing streaming scenarios and keeping the
working set which can fit in the LLC. An example bypassing
technique performs random bypassing of the cache lines based on a
probability which is increased or decreased based on the references
to the bypassed lines. This bypassing technique utilizes an
additional tag structure to store the tag of the bypassed line and
a pointer to the replacement victim which would have been evicted
without bypassing. Any suitable technology may be utilized to
identify bypass candidates, including re-use-count,
re-use-distance, etc.
[0071] Because bypassing all requests degrades performance, some
cache systems may utilize adaptive bypassing that performs bypass
only if no invalid blocks are available in LLC. For exclusive LLC,
such systems may include bypass and insertion age techniques. The
LLC bypass and age assignment decisions may be based on two
properties of the data line when it is considered for allocation in
the LLC. The first property is the number of trips (trip count)
made by the data line between the MLC and the LLC from the time it
is brought into the cache hierarchy till it is evicted from the
LLC. The second property is the number of MLC cache hits (use
count) experienced by a data line during its residency in the MLC.
For each category of use count and trip count (e.g., which may
collectively be referred as a dead block prediction (DBP) bin), a
DBP module may maintain a LLC hit rate counter for some of the
sample sets (e.g., which may be referred to as "observer sets").
For example, sampling may be performed only for the few sets to
reduce the overhead of the cache profiling. When there is an MLC
eviction belonging to a certain category of the DBP Bin for the
non-observer sets (e.g., also referred to as "follower sets"), the
DBP module checks the corresponding LLC hit rate counter for this
category in the observer set. When the LLC hit rate is less than a
configurable threshold, then the DBP module may determine that the
probability of this line getting re-used from LLC is lower and may
mark the line as "dead" before sending it to the LLC. When the LLC
receives a "dead" eviction, the cache controller may insert the
line at LRU in the LLC if an invalid block is available in LLC,
otherwise the cache controller bypasses the LLC. Inserting the line
at LRU ensures that the line becomes a victim candidate first
before evicting existing non-LRU lines in the LLC.
[0072] Some embodiments may focus on a specific DBP bin, which may
be referred to as single re-use, that corresponds to a use count
value of one (1) and a trip count value of zero (0). In accordance
with some embodiments, a single re-use data line is read from the
main memory (e.g., either directly as a core demand or MLC
pre-fetch, or pre-fetched into the LLC as an LLC pre-fetch and then
read from the LLC) and is accessed exactly once in the MLC.
[0073] As noted above, DBP technology may utilize observer sets to
detect a streaming scenario. The observer sets provide an
indication if there is a re-use of the lines from the LLC. The core
then tries to prevent thrashing from the follower sets. For the
example "Streaming" scenario from FIG. 10, because there is almost
no re-use from the LLC, the core learns about this streaming
pattern from the observer set and then bypasses the LLC for the
"A", "B", "C" and "D" accesses. With the bypass, the footprint that
was in the LLC prior to the streaming access may be retained. The
bypass may be achieved by installing the MLC evictions of the
streaming buffers at LRU. The LRU lines become the candidates for
the next LLC eviction thereby preserving the existing lines in the
LLC which might see a future re-use.
[0074] For the example "Single Re-Use" scenario from FIG. 10, DBP
technology detects "D1" as streaming and hence bypasses "D1" for
the follower. The "C1" buffer is treated in the same way as the
buffer "D1". The re-use of the buffer "C1" is seen only in "C2" in
the observer sets. Conventionally, however, "C1" is entirely
bypassed for the follower sets and accordingly "C2" turns out to be
an LLC Miss though the capacity fits in the LLC. Single re-use data
presents a problem for conventional non-inclusive LLC with DBP,
because there is no way to predict the capacity of the incoming
buffer. The DBP learns the re-use only when it observes the second
iteration of the buffer access in the observer set, which is too
late, and the buffer is already bypassed for the first iteration in
the follower sets. For the example "Multi Re-Use" scenario in FIG.
10, the observer set learns about the re-use during the second
iteration and now installs the buffer at a higher age for the
follower sets, which ensures that the third and subsequent re-use
is captured in the LLC.
[0075] Conventionally, for an MLC eviction which DBP marks as dead,
the line is installed in LRU if an invalid block is available in
the LLC to prevent future thrashing (e.g., because the line itself
becomes the first candidate for eviction from the LLC). With the
line in LRU, however, there is a chance of getting an opportunistic
LLC hit before getting evicted. While this technique works with
most of the DBP bins, this technique cannot capture the "Single
Re-Use" scenario. The moment the application accesses a new buffer
during the first iteration, DBP marks all evictions as "dead" until
it starts achieving re-use for the second iteration in the observer
sets.
[0076] As noted above, a cache controller may maintain a LHC which
is incremented on every LLC hit to an earlier MLC eviction in the
LLC. In order to solve the problem of capturing single re-use data,
some embodiments may check the global LLC hit rate in the observer
set across all DBP bins. The global LLC hit rate in the observer
set across all DBP bins may indicate from history if the
application has seen any kind of re-use from the LLC independent of
the DBP bin. Some embodiments of SRP technology may check the
following two parameters: A) if the LHC is greater than a
threshold, signifying re-use from the LLC; and B) if the origin of
the request was main memory. When both these conditions are met,
the data may be installed from the MLC in MRU instead of LRU.
Embodiments identify main memory as the source of a request that
may potentially observe future re-use, which is not captured by
conventional DBP techniques for non-inclusive LLC.
[0077] With reference to FIG. 11, an embodiment of a method 600 of
controlling a cache may start with an MLC eviction of a data line
at box 631 that DBP marks as dead at box 632. The method 600 may
then include determining if LHC is greater than a threshold value
and if a source of the data line is main memory at box 633. If the
two conditions are both met at box 633, the method 600 may proceed
to installing the line at MRU in the LLC at box 634. If the two
conditions are not both met at box 633, the method 600 may proceed
to determining if an invalid block is available at box 635 and, if
so, installing the line at LRU in the LLC at box 636. Otherwise the
method 600 may proceed to bypassing the LLC at box 637.
[0078] With reference to FIG. 12, an embodiment of a cache system
700 includes a cache controller 712 communicatively coupled to a
core cache 714, and a LLC 716. The core cache 714 includes a L1
cache 714a and a MLC 714b). The cache controller 712 maintains a
LHC which is incremented on every LLC hit to an earlier MLC
eviction in the LLC 716. The cache controller 712 is configured
with SRP technology to handle an eviction of a data line from the
MLC 714b that is marked as dead as follows. The cache controller
712 determines if LHC is greater than a threshold value and if a
source of the data line is main memory. If the two conditions are
both met, the cache controller 712 proceeds to install the line at
MRU in the LLC 716. If the two conditions are not both, the cache
controller 712 proceeds to determine if an invalid block is
available in the LLC 716 and, if so, the cache controller 712
installs the line at LRU in the LLC 716. Otherwise, if an invalid
block is not available in the LLC 716, the cache controller 712
proceeds to bypass the LLC 716.
[0079] Performance modeling of embodiments of SRP technology in a
cycle accurate model shows increased LLC hit rate for single re-use
data, increased instructions per cycle (IPC), and reduced memory
access (improved bandwidth), as compared to a baseline
non-exclusive LLC without SRP technology.
[0080] Those skilled in the art will appreciate that a wide variety
of devices may benefit from the foregoing embodiments. The
following exemplary core architectures, processors, and computer
architectures are non-limiting examples of devices that may
beneficially incorporate embodiments of the technology described
herein.
[0081] Exemplary Core Architectures, Processors, and Computer
Architectures
[0082] Processor cores may be implemented in different ways, for
different purposes, and in different processors. For instance,
implementations of such cores may include: 1) a general purpose
in-order core intended for general-purpose computing; 2) a high
performance general purpose out-of-order core intended for
general-purpose computing; 3) a special purpose core intended
primarily for graphics and/or scientific (throughput) computing.
Implementations of different processors may include: 1) a CPU
including one or more general purpose in-order cores intended for
general-purpose computing and/or one or more general purpose
out-of-order cores intended for general-purpose computing; and 2) a
coprocessor including one or more special purpose cores intended
primarily for graphics and/or scientific (throughput). Such
different processors lead to different computer system
architectures, which may include: 1) the coprocessor on a separate
chip from the CPU; 2) the coprocessor on a separate die in the same
package as a CPU; 3) the coprocessor on the same die as a CPU (in
which case, such a coprocessor is sometimes referred to as special
purpose logic, such as integrated graphics and/or scientific
(throughput) logic, or as special purpose cores); and 4) a system
on a chip that may include on the same die the described CPU
(sometimes referred to as the application core(s) or application
processor(s)), the above described coprocessor, and additional
functionality. Exemplary core architectures are described next,
followed by descriptions of exemplary processors and computer
architectures.
[0083] Exemplary Core Architectures
[0084] In-Order and Out-of-Order Core Block Diagram
[0085] FIG. 13A is a block diagram illustrating both an exemplary
in-order pipeline and an exemplary register renaming, out-of-order
issue/execution pipeline according to embodiments of the invention.
FIG. 13B is a block diagram illustrating both an exemplary
embodiment of an in-order architecture core and an exemplary
register renaming, out-of-order issue/execution architecture core
to be included in a processor according to embodiments of the
invention. The solid lined boxes in FIGS. 13A-B illustrate the
in-order pipeline and in-order core, while the optional addition of
the dashed lined boxes illustrates the register renaming,
out-of-order issue/execution pipeline and core. Given that the
in-order aspect is a subset of the out-of-order aspect, the
out-of-order aspect will be described.
[0086] In FIG. 13A, a processor pipeline 900 includes a fetch stage
902, a length decode stage 904, a decode stage 906, an allocation
stage 908, a renaming stage 910, a scheduling (also known as a
dispatch or issue) stage 912, a register read/memory read stage
914, an execute stage 916, a write back/memory write stage 918, an
exception handling stage 922, and a commit stage 924.
[0087] FIG. 13B shows processor core 990 including a front end unit
930 coupled to an execution engine unit 950, and both are coupled
to a memory unit 970. The core 990 may be a reduced instruction set
computing (RISC) core, a complex instruction set computing (CISC)
core, a very long instruction word (VLIW) core, or a hybrid or
alternative core type. As yet another option, the core 990 may be a
special-purpose core, such as, for example, a network or
communication core, compression engine, coprocessor core, general
purpose computing graphics processing unit (GPGPU) core, graphics
core, or the like.
[0088] The front end unit 930 includes a branch prediction unit 932
coupled to an instruction cache unit 934, which is coupled to an
instruction translation lookaside buffer (TLB) 936, which is
coupled to an instruction fetch unit 938, which is coupled to a
decode unit 940. The decode unit 940 (or decoder) may decode
instructions, and generate as an output one or more
micro-operations, micro-code entry points, microinstructions, other
instructions, or other control signals, which are decoded from, or
which otherwise reflect, or are derived from, the original
instructions. The decode unit 940 may be implemented using various
different mechanisms. Examples of suitable mechanisms include, but
are not limited to, look-up tables, hardware implementations,
programmable logic arrays (PLAs), microcode read only memories
(ROMs), etc. In one embodiment, the core 990 includes a microcode
ROM or other medium that stores microcode for certain
macroinstructions (e.g., in decode unit 940 or otherwise within the
front end unit 930). The decode unit 940 is coupled to a
rename/allocator unit 952 in the execution engine unit 950.
[0089] The execution engine unit 950 includes the rename/allocator
unit 952 coupled to a retirement unit 954 and a set of one or more
scheduler unit(s) 956. The scheduler unit(s) 956 represents any
number of different schedulers, including reservations stations,
central instruction window, etc. The scheduler unit(s) 956 is
coupled to the physical register file(s) unit(s) 958. Each of the
physical register file(s) units 958 represents one or more physical
register files, different ones of which store one or more different
data types, such as scalar integer, scalar floating point, packed
integer, packed floating point, vector integer, vector floating
point, status (e.g., an instruction pointer that is the address of
the next instruction to be executed), etc. In one embodiment, the
physical register file(s) unit 958 comprises a vector registers
unit, a write mask registers unit, and a scalar registers unit.
These register units may provide architectural vector registers,
vector mask registers, and general purpose registers. The physical
register file(s) unit(s) 958 is overlapped by the retirement unit
954 to illustrate various ways in which register renaming and
out-of-order execution may be implemented (e.g., using a reorder
buffer(s) and a retirement register file(s); using a future
file(s), a history buffer(s), and a retirement register file(s);
using a register maps and a pool of registers; etc.). The
retirement unit 954 and the physical register file(s) unit(s) 958
are coupled to the execution cluster(s) 960. The execution
cluster(s) 960 includes a set of one or more execution units 962
and a set of one or more memory access units 964. The execution
units 962 may perform various operations (e.g., shifts, addition,
subtraction, multiplication) and on various types of data (e.g.,
scalar floating point, packed integer, packed floating point,
vector integer, vector floating point). While some embodiments may
include a number of execution units dedicated to specific functions
or sets of functions, other embodiments may include only one
execution unit or multiple execution units that all perform all
functions. The scheduler unit(s) 956, physical register file(s)
unit(s) 958, and execution cluster(s) 960 are shown as being
possibly plural because certain embodiments create separate
pipelines for certain types of data/operations (e.g., a scalar
integer pipeline, a scalar floating point/packed integer/packed
floating point/vector integer/vector floating point pipeline,
and/or a memory access pipeline that each have their own scheduler
unit, physical register file(s) unit, and/or execution cluster--and
in the case of a separate memory access pipeline, certain
embodiments are implemented in which only the execution cluster of
this pipeline has the memory access unit(s) 964). It should also be
understood that where separate pipelines are used, one or more of
these pipelines may be out-of-order issue/execution and the rest
in-order.
[0090] The set of memory access units 964 is coupled to the memory
unit 970, which includes a data TLB unit 972 coupled to a data
cache unit 974 coupled to a level 2 (L2) cache unit 976. In one
exemplary embodiment, the memory access units 964 may include a
load unit, a store address unit, and a store data unit, each of
which is coupled to the data TLB unit 972 in the memory unit 970.
The instruction cache unit 934 is further coupled to a level 2 (L2)
cache unit 976 in the memory unit 970. The L2 cache unit 976 is
coupled to one or more other levels of cache and eventually to a
main memory.
[0091] By way of example, the exemplary register renaming,
out-of-order issue/execution core architecture may implement the
pipeline 900 as follows: 1) the instruction fetch 938 performs the
fetch and length decoding stages 902 and 904; 2) the decode unit
940 performs the decode stage 906; 3) the rename/allocator unit 952
performs the allocation stage 908 and renaming stage 910; 4) the
scheduler unit(s) 956 performs the schedule stage 912; 5) the
physical register file(s) unit(s) 958 and the memory unit 970
perform the register read/memory read stage 914; the execution
cluster 960 perform the execute stage 916; 6) the memory unit 970
and the physical register file(s) unit(s) 958 perform the write
back/memory write stage 918; 7) various units may be involved in
the exception handling stage 922; and 8) the retirement unit 954
and the physical register file(s) unit(s) 958 perform the commit
stage 924.
[0092] The core 990 may support one or more instructions sets
(e.g., the x86 instruction set (with some extensions that have been
added with newer versions); the MIPS instruction set of MIPS
Technologies of Sunnyvale, Calif.; the ARM instruction set (with
optional additional extensions such as NEON) of ARM Holdings of
Sunnyvale, Calif.), including the instruction(s) described herein.
In one embodiment, the core 990 includes logic to support a packed
data instruction set extension (e.g., AVX1, AVX2), thereby allowing
the operations used by many multimedia applications to be performed
using packed data.
[0093] It should be understood that the core may support
multithreading (executing two or more parallel sets of operations
or threads), and may do so in a variety of ways including time
sliced multithreading, simultaneous multithreading (where a single
physical core provides a logical core for each of the threads that
physical core is simultaneously multithreading), or a combination
thereof (e.g., time sliced fetching and decoding and simultaneous
multithreading thereafter such as in the Intel.RTM. Hyperthreading
technology).
[0094] While register renaming is described in the context of
out-of-order execution, it should be understood that register
renaming may be used in an in-order architecture. While the
illustrated embodiment of the processor also includes separate
instruction and data cache units 934/974 and a shared L2 cache unit
976, alternative embodiments may have a single internal cache for
both instructions and data, such as, for example, a Level 1 (L1)
internal cache, or multiple levels of internal cache. In some
embodiments, the system may include a combination of an internal
cache and an external cache that is external to the core and/or the
processor. Alternatively, all of the cache may be external to the
core and/or the processor.
[0095] Specific Exemplary In-Order Core Architecture
[0096] FIGS. 14A-B illustrate a block diagram of a more specific
exemplary in-order core architecture, which core would be one of
several logic blocks (including other cores of the same type and/or
different types) in a chip. The logic blocks communicate through a
high-bandwidth interconnect network (e.g., a ring network) with
some fixed function logic, memory I/O interfaces, and other
necessary I/O logic, depending on the application.
[0097] FIG. 14A is a block diagram of a single processor core,
along with its connection to the on-die interconnect network 1002
and with its local subset of the Level 2 (L2) cache 1004, according
to embodiments of the invention. In one embodiment, an instruction
decoder 1000 supports the x86 instruction set with a packed data
instruction set extension. An L1 cache 1006 allows low-latency
accesses to cache memory into the scalar and vector units. While in
one embodiment (to simplify the design), a scalar unit 1008 and a
vector unit 1010 use separate register sets (respectively, scalar
registers 1012 and vector registers 1014) and data transferred
between them is written to memory and then read back in from a
level 1 (L1) cache 1006, alternative embodiments of the invention
may use a different approach (e.g., use a single register set or
include a communication path that allow data to be transferred
between the two register files without being written and read
back).
[0098] The local subset of the L2 cache 1004 is part of a global L2
cache that is divided into separate local subsets, one per
processor core. Each processor core has a direct access path to its
own local subset of the L2 cache 1004. Data read by a processor
core is stored in its L2 cache subset 1004 and can be accessed
quickly, in parallel with other processor cores accessing their own
local L2 cache subsets. Data written by a processor core is stored
in its own L2 cache subset 1004 and is flushed from other subsets,
if necessary. The ring network ensures coherency for shared data.
The ring network is bi-directional to allow agents such as
processor cores, L2 caches and other logic blocks to communicate
with each other within the chip. Each ring data-path is 1012-bits
wide per direction.
[0099] FIG. 14B is an expanded view of part of the processor core
in FIG. 14A according to embodiments of the invention. FIG. 14B
includes an L1 data cache 1006A part of the L1 cache 1006, as well
as more detail regarding the vector unit 1010 and the vector
registers 1014. Specifically, the vector unit 1010 is a 16-wide
vector processing unit (VPU) (see the 16-wide ALU 1028), which
executes one or more of integer, single-precision float, and
double-precision float instructions. The VPU supports swizzling the
register inputs with swizzle unit 1020, numeric conversion with
numeric convert units 1022A-B, and replication with replication
unit 1024 on the memory input. Write mask registers 1026 allow
predicating resulting vector writes.
[0100] FIG. 15 is a block diagram of a processor 1100 that may have
more than one core, may have an integrated memory controller, and
may have integrated graphics according to embodiments of the
invention. The solid lined boxes in FIG. 15 illustrate a processor
1100 with a single core 1102A, a system agent 1110, a set of one or
more bus controller units 1116, while the optional addition of the
dashed lined boxes illustrates an alternative processor 1100 with
multiple cores 1102A-N, a set of one or more integrated memory
controller unit(s) 1114 in the system agent unit 1110, and special
purpose logic 1108.
[0101] Thus, different implementations of the processor 1100 may
include: 1) a CPU with the special purpose logic 1108 being
integrated graphics and/or scientific (throughput) logic (which may
include one or more cores), and the cores 1102A-N being one or more
general purpose cores (e.g., general purpose in-order cores,
general purpose out-of-order cores, a combination of the two); 2) a
coprocessor with the cores 1102A-N being a large number of special
purpose cores intended primarily for graphics and/or scientific
(throughput); and 3) a coprocessor with the cores 1102A-N being a
large number of general purpose in-order cores. Thus, the processor
1100 may be a general-purpose processor, coprocessor or
special-purpose processor, such as, for example, a network or
communication processor, compression engine, graphics processor,
GPGPU (general purpose graphics processing unit), a high-throughput
many integrated core (MIC) coprocessor (including 30 or more
cores), embedded processor, or the like. The processor may be
implemented on one or more chips. The processor 1100 may be a part
of and/or may be implemented on one or more substrates using any of
a number of process technologies, such as, for example, BiCMOS,
CMOS, or NMOS.
[0102] The memory hierarchy includes one or more levels of
respective caches 1104A-N within the cores 1102A-N, a set or one or
more shared cache units 1106, and external memory (not shown)
coupled to the set of integrated memory controller units 1114. The
set of shared cache units 1106 may include one or more mid-level
caches, such as level 2 (L2), level 3 (L3), level 4 (L4), or other
levels of cache, a last level cache (LLC), and/or combinations
thereof. While in one embodiment a ring based interconnect unit
1112 interconnects the integrated graphics logic 1108, the set of
shared cache units 1106, and the system agent unit 1110/integrated
memory controller unit(s) 1114, alternative embodiments may use any
number of well-known techniques for interconnecting such units. In
one embodiment, coherency is maintained between one or more cache
units 1106 and cores 1102-A-N.
[0103] In some embodiments, one or more of the cores 1102A-N are
capable of multithreading. The system agent 1110 includes those
components coordinating and operating cores 1102A-N. The system
agent unit 1110 may include for example a power control unit (PCU)
and a display unit. The PCU may be or include logic and components
needed for regulating the power state of the cores 1102A-N and the
integrated graphics logic 1108. The display unit is for driving one
or more externally connected displays.
[0104] The cores 1102A-N may be homogenous or heterogeneous in
terms of architecture instruction set; that is, two or more of the
cores 1102A-N may be capable of execution the same instruction set,
while others may be capable of executing only a subset of that
instruction set or a different instruction set.
[0105] Exemplary Computer Architectures
[0106] FIGS. 16-19 are block diagrams of exemplary computer
architectures. Other system designs and configurations known in the
arts for laptops, desktops, handheld PCs, personal digital
assistants, engineering workstations, servers, network devices,
network hubs, switches, embedded processors, digital signal
processors (DSPs), graphics devices, video game devices, set-top
boxes, micro controllers, cell phones, portable media players, hand
held devices, and various other electronic devices, are also
suitable. In general, a huge variety of systems or electronic
devices capable of incorporating a processor and/or other execution
logic as disclosed herein are generally suitable.
[0107] Referring now to FIG. 16, shown is a block diagram of a
system 1200 in accordance with one embodiment of the present
invention. The system 1200 may include one or more processors 1210,
1215, which are coupled to a controller hub 1220. In one embodiment
the controller hub 1220 includes a graphics memory controller hub
(GMCH) 1290 and an Input/Output Hub (IOH) 1250 (which may be on
separate chips); the GMCH 1290 includes memory and graphics
controllers to which are coupled memory 1240 and a coprocessor
1245; the IOH 1250 couples input/output (I/O) devices 1260 to the
GMCH 1290. Alternatively, one or both of the memory and graphics
controllers are integrated within the processor (as described
herein), the memory 1240 and the coprocessor 1245 are coupled
directly to the processor 1210, and the controller hub 1220 in a
single chip with the IOH 1250.
[0108] The optional nature of additional processors 1215 is denoted
in FIG. 16 with broken lines. Each processor 1210, 1215 may include
one or more of the processing cores described herein and may be
some version of the processor 1100.
[0109] The memory 1240 may be, for example, dynamic random access
memory (DRAM), phase change memory (PCM), or a combination of the
two. For at least one embodiment, the controller hub 1220
communicates with the processor(s) 1210, 1215 via a multi-drop bus,
such as a frontside bus (FSB), point-to-point interface such as
QuickPath Interconnect (QPI), or similar connection 1295.
[0110] In one embodiment, the coprocessor 1245 is a special-purpose
processor, such as, for example, a high-throughput MIC processor, a
network or communication processor, compression engine, graphics
processor, GPGPU, embedded processor, or the like. In one
embodiment, controller hub 1220 may include an integrated graphics
accelerator.
[0111] There can be a variety of differences between the physical
resources 1210, 1215 in terms of a spectrum of metrics of merit
including architectural, microarchitectural, thermal, power
consumption characteristics, and the like.
[0112] In one embodiment, the processor 1210 executes instructions
that control data processing operations of a general type. Embedded
within the instructions may be coprocessor instructions. The
processor 1210 recognizes these coprocessor instructions as being
of a type that should be executed by the attached coprocessor 1245.
Accordingly, the processor 1210 issues these coprocessor
instructions (or control signals representing coprocessor
instructions) on a coprocessor bus or other interconnect, to
coprocessor 1245. Coprocessor(s) 1245 accept and execute the
received coprocessor instructions.
[0113] Referring now to FIG. 17, shown is a block diagram of a
first more specific exemplary system 1300 in accordance with an
embodiment of the present invention. As shown in FIG. 17,
multiprocessor system 1300 is a point-to-point interconnect system,
and includes a first processor 1370 and a second processor 1380
coupled via a point-to-point interconnect 1350. Each of processors
1370 and 1380 may be some version of the processor 1100. In one
embodiment of the invention, processors 1370 and 1380 are
respectively processors 1210 and 1215, while coprocessor 1338 is
coprocessor 1245. In another embodiment, processors 1370 and 1380
are respectively processor 1210 coprocessor 1245.
[0114] Processors 1370 and 1380 are shown including integrated
memory controller (IMC) units 1372 and 1382, respectively.
Processor 1370 also includes as part of its bus controller units
point-to-point (P-P) interfaces 1376 and 1378; similarly, second
processor 1380 includes P-P interfaces 1386 and 1388. Processors
1370, 1380 may exchange information via a point-to-point (P-P)
interface 1350 using P-P interface circuits 1378, 1388. As shown in
FIG. 17, IMCs 1372 and 1382 couple the processors to respective
memories, namely a memory 1332 and a memory 1334, which may be
portions of main memory locally attached to the respective
processors.
[0115] Processors 1370, 1380 may each exchange information with a
chipset 1390 via individual P-P interfaces 1352, 1354 using point
to point interface circuits 1376, 1394, 1386, 1398. Chipset 1390
may optionally exchange information with the coprocessor 1338 via a
high-performance interface 1339 and an interface 1392. In one
embodiment, the coprocessor 1338 is a special-purpose processor,
such as, for example, a high-throughput MIC processor, a network or
communication processor, compression engine, graphics processor,
GPGPU, embedded processor, or the like.
[0116] A shared cache (not shown) may be included in either
processor or outside of both processors, yet connected with the
processors via P-P interconnect, such that either or both
processors' local cache information may be stored in the shared
cache if a processor is placed into a low power mode.
[0117] Chipset 1390 may be coupled to a first bus 1316 via an
interface 1396. In one embodiment, first bus 1316 may be a
Peripheral Component Interconnect (PCI) bus, or a bus such as a PCI
Express bus or another third generation I/O interconnect bus,
although the scope of the present invention is not so limited.
[0118] As shown in FIG. 17, various I/O devices 1314 may be coupled
to first bus 1316, along with a bus bridge 1318 which couples first
bus 1316 to a second bus 1320. In one embodiment, one or more
additional processor(s) 1315, such as coprocessors, high-throughput
MIC processors, GPGPU's, accelerators (such as, e.g., graphics
accelerators or digital signal processing (DSP) units), field
programmable gate arrays, or any other processor, are coupled to
first bus 1316. In one embodiment, second bus 1320 may be a low pin
count (LPC) bus. Various devices may be coupled to a second bus
1320 including, for example, a keyboard and/or mouse 1322,
communication devices 1327 and a storage unit 1328 such as a disk
drive or other mass storage device which may include
instructions/code and data 1330, in one embodiment. Further, an
audio I/O 1324 may be coupled to the second bus 1320. Note that
other architectures are possible. For example, instead of the
point-to-point architecture of FIG. 17, a system may implement a
multi-drop bus or other such architecture.
[0119] Referring now to FIG. 18, shown is a block diagram of a
second more specific exemplary system 1400 in accordance with an
embodiment of the present invention Like elements in FIGS. 17 and
18 bear like reference numerals, and certain aspects of FIG. 17
have been omitted from FIG. 18 in order to avoid obscuring other
aspects of FIG. 18.
[0120] FIG. 18 illustrates that the processors 1370, 1380 may
include integrated memory and I/O control logic ("CL") 1472 and
1482, respectively. Thus, the CL 1472, 1482 include integrated
memory controller units and include I/O control logic. FIG. 18
illustrates that not only are the memories 1332, 1334 coupled to
the CL 1472, 1482, but also that I/O devices 1414 are also coupled
to the control logic 1472, 1482. Legacy I/O devices 1415 are
coupled to the chipset 1390.
[0121] Referring now to FIG. 19, shown is a block diagram of a SoC
1500 in accordance with an embodiment of the present invention.
Similar elements in FIG. 15 bear like reference numerals. Also,
dashed lined boxes are optional features on more advanced SoCs. In
FIG. 19, an interconnect unit(s) 1502 is coupled to: an application
processor 1510 which includes a set of one or more cores 1102A-N
and shared cache unit(s) 1106; a system agent unit 1110; a bus
controller unit(s) 1116; an integrated memory controller unit(s)
1114; a set or one or more coprocessors 1520 which may include
integrated graphics logic, an image processor, an audio processor,
and a video processor; an static random access memory (SRAM) unit
1530; a direct memory access (DMA) unit 1532; and a display unit
1540 for coupling to one or more external displays. In one
embodiment, the coprocessor(s) 1520 include a special-purpose
processor, such as, for example, a network or communication
processor, compression engine, GPGPU, a high-throughput MIC
processor, embedded processor, or the like.
[0122] Embodiments of the mechanisms disclosed herein may be
implemented in hardware, software, firmware, or a combination of
such implementation approaches. Embodiments of the invention may be
implemented as computer programs or program code executing on
programmable systems comprising at least one processor, a storage
system (including volatile and non-volatile memory and/or storage
elements), at least one input device, and at least one output
device.
[0123] Program code, such as code 1330 illustrated in FIG. 17, may
be applied to input instructions to perform the functions described
herein and generate output information. The output information may
be applied to one or more output devices, in known fashion. For
purposes of this application, a processing system includes any
system that has a processor, such as, for example; a digital signal
processor (DSP), a microcontroller, an application specific
integrated circuit (ASIC), or a microprocessor.
[0124] The program code may be implemented in a high level
procedural or object oriented programming language to communicate
with a processing system. The program code may also be implemented
in assembly or machine language, if desired. In fact, the
mechanisms described herein are not limited in scope to any
particular programming language. In any case, the language may be a
compiled or interpreted language.
[0125] One or more aspects of at least one embodiment may be
implemented by representative instructions stored on a
machine-readable medium which represents various logic within the
processor, which when read by a machine causes the machine to
fabricate logic to perform the techniques described herein. Such
representations, known as "IP cores" may be stored on a tangible,
machine readable medium and supplied to various customers or
manufacturing facilities to load into the fabrication machines that
actually make the logic or processor.
[0126] Such machine-readable storage media may include, without
limitation, non-transitory, tangible arrangements of articles
manufactured or formed by a machine or device, including storage
media such as hard disks, any other type of disk including floppy
disks, optical disks, compact disk read-only memories (CD-ROMs),
compact disk rewritable's (CD-RWs), and magneto-optical disks,
semiconductor devices such as read-only memories (ROMs), random
access memories (RAMs) such as dynamic random access memories
(DRAMs), static random access memories (SRAMs), erasable
programmable read-only memories (EPROMs), flash memories,
electrically erasable programmable read-only memories (EEPROMs),
phase change memory (PCM), magnetic or optical cards, or any other
type of media suitable for storing electronic instructions.
[0127] Accordingly, embodiments of the invention also include
non-transitory, tangible machine-readable media containing
instructions or containing design data, such as Hardware
Description Language (HDL), which defines structures, circuits,
apparatuses, processors and/or system features described herein.
Such embodiments may also be referred to as program products.
[0128] Emulation (Including Binary Translation, Code Morphing,
Etc.)
[0129] In some cases, an instruction converter may be used to
convert an instruction from a source instruction set to a target
instruction set. For example, the instruction converter may
translate (e.g., using static binary translation, dynamic binary
translation including dynamic compilation), morph, emulate, or
otherwise convert an instruction to one or more other instructions
to be processed by the core. The instruction converter may be
implemented in software, hardware, firmware, or a combination
thereof. The instruction converter may be on processor, off
processor, or part on and part off processor.
[0130] FIG. 20 is a block diagram contrasting the use of a software
instruction converter to convert binary instructions in a source
instruction set to binary instructions in a target instruction set
according to embodiments of the invention. In the illustrated
embodiment, the instruction converter is a software instruction
converter, although alternatively the instruction converter may be
implemented in software, firmware, hardware, or various
combinations thereof. FIG. 20 shows a program in a high level
language 1602 may be compiled using an x86 compiler 1604 to
generate x86 binary code 1606 that may be natively executed by a
processor with at least one x86 instruction set core 1616. The
processor with at least one x86 instruction set core 1616
represents any processor that can perform substantially the same
functions as an Intel processor with at least one x86 instruction
set core by compatibly executing or otherwise processing (1) a
substantial portion of the instruction set of the Intel x86
instruction set core or (2) object code versions of applications or
other software targeted to run on an Intel processor with at least
one x86 instruction set core, in order to achieve substantially the
same result as an Intel processor with at least one x86 instruction
set core. The x86 compiler 1604 represents a compiler that is
operable to generate x86 binary code 1606 (e.g., object code) that
can, with or without additional linkage processing, be executed on
the processor with at least one x86 instruction set core 1616.
Similarly, FIG. 20 shows the program in the high level language
1602 may be compiled using an alternative instruction set compiler
1608 to generate alternative instruction set binary code 1610 that
may be natively executed by a processor without at least one x86
instruction set core 1614 (e.g., a processor with cores that
execute the MIPS instruction set of MIPS Technologies of Sunnyvale,
Calif. and/or that execute the ARM instruction set of ARM Holdings
of Sunnyvale, Calif.). The instruction converter 1612 is used to
convert the x86 binary code 1606 into code that may be natively
executed by the processor without an x86 instruction set core 1614.
This converted code is not likely to be the same as the alternative
instruction set binary code 1610 because an instruction converter
capable of this is difficult to make; however, the converted code
will accomplish the general operation and be made up of
instructions from the alternative instruction set. Thus, the
instruction converter 1612 represents software, firmware, hardware,
or a combination thereof that, through emulation, simulation or any
other process, allows a processor or other electronic device that
does not have an x86 instruction set processor or core to execute
the x86 binary code 1606.
[0131] Techniques and architectures for instruction set
architecture opcode parameterization are described herein. In the
above description, for purposes of explanation, numerous specific
details are set forth in order to provide a thorough understanding
of certain embodiments. It will be apparent, however, to one
skilled in the art that certain embodiments can be practiced
without these specific details. In other instances, structures and
devices are shown in block diagram form in order to avoid obscuring
the description.
ADDITIONAL NOTES AND EXAMPLES
[0132] Example 1 includes an integrated circuit, comprising a core,
and a cache controller coupled to the core, the cache controller
including circuitry to identify data from a working set for dynamic
inclusion in a next level cache based on an amount of re-use of the
next level cache, send a shared copy of the identified data to a
requesting core of one or more processor cores, and maintain a copy
of the identified data in the next level cache.
[0133] Example 2 includes the integrated circuit of claim 1,
wherein the circuitry is further to determine dynamic inclusion of
data in the next level cache on a per data line basis.
[0134] Example 3 includes the integrated circuit of claim 1,
wherein the circuitry is further to increment a counter value when
a hit in the next level cache corresponds to an eviction from a
core cache, and identify a current data hit in the next level cache
for dynamic inclusion in the next level cache if the current data
hit corresponds to an eviction from the core cache and if the
counter value is greater than a threshold.
[0135] Example 4 includes the integrated circuit of claim 3,
wherein the circuitry is further to set a snoop filter to indicate
that the requesting core is valid for the current data hit.
[0136] Example 5 includes the integrated circuit of claim 4,
wherein, if the current data hit does not correspond to an eviction
from the core cache or if the counter value is not greater than the
threshold, the circuitry is further to send an exclusive copy of
the data to the requesting core, update an entry in the snoop
filter to indicate a core identifier of the requesting core, and
deallocate the data in the next level cache.
[0137] Example 6 includes the integrated circuit of claim 1,
wherein the circuitry is further to silently drop data to be
evicted from a core cache if the data to be evicted from the core
cache has a shared copy of the data in the next level cache.
[0138] Example 7 includes the integrated circuit of claim 1,
wherein the next level cache comprises a non-inclusive last level
cache.
[0139] Example 8 includes a method of controlling a cache,
comprising identifying data from a working set for dynamic
inclusion in a next level cache based on an amount of re-use of the
next level cache, sending a shared copy of the identified data to a
requesting core of one or more processor cores, and maintaining a
copy of the identified data in the next level cache.
[0140] Example 9 includes the method of claim 8, further comprising
determining dynamic inclusion of data in the next level cache on a
per data line basis.
[0141] Example 10 includes the method of claim 8, further
comprising incrementing a counter value when a hit in the next
level cache corresponds to an eviction from a core cache, and
identifying a current data hit in the next level cache for dynamic
inclusion in the next level cache if the current data hit
corresponds to an eviction from the core cache and if the counter
value is greater than a threshold.
[0142] Example 11 includes the method of claim 10, further
comprising setting a snoop filter to indicate that the requesting
core is valid for the current data hit.
[0143] Example 12 includes the method of claim 11, wherein, if the
current data hit does not correspond to an eviction from the core
cache or if the counter value is not greater than the threshold,
the method further comprises sending an exclusive copy of the data
to the requesting core, updating an entry in the snoop filter to
indicate a core identifier of the requesting core, and deallocating
the data in the next level cache.
[0144] Example 13 includes the method of claim 8, further
comprising silently dropping data to be evicted from a core cache
if the data to be evicted from the core cache has a shared copy of
the data in the next level cache.
[0145] Example 14 includes an apparatus, comprising one or more
processor cores, a core cache co-located with and communicatively
coupled to the one or more processor cores, a next level cache
co-located with and communicatively coupled to the core cache and
the one or more processor cores, and a cache controller co-located
with and communicatively coupled to the core cache, the next level
cache, and the one or more processor cores, the cache controller
including circuitry to identify data from a working set for dynamic
inclusion in the next level cache based on an amount of re-use of
the next level cache, send a shared copy of the identified data to
a requesting core of the one or more processor cores, and maintain
a copy of the identified data in the next level cache.
[0146] Example 15 includes the apparatus of claim 14, wherein the
circuitry is further to determine dynamic inclusion of data in the
next level cache on a per data line basis.
[0147] Example 16 includes the apparatus of claim 14, wherein the
circuitry is further to increment a counter value when a hit in the
next level cache corresponds to an eviction from the core cache,
and identify a current data hit in the next level cache for dynamic
inclusion in the next level cache if the current data hit
corresponds to an eviction from the core cache and if the counter
value is greater than a threshold.
[0148] Example 17 includes the apparatus of claim 16, wherein the
circuitry is further to set a snoop filter to indicate that the
requesting core is valid for the current data hit.
[0149] Example 18 includes the apparatus of claim 16, wherein, if
the current data hit does not correspond to an eviction from the
core cache or if the counter value is not greater than the
threshold, the circuitry is further to send an exclusive copy of
the data to the requesting core, update an entry in the snoop
filter to indicate a core identifier of the requesting core, and
deallocate the data in the next level cache.
[0150] Example 19 includes the apparatus of claim 14, wherein the
circuitry is further to silently drop data to be evicted from a
core cache if the data to be evicted from the core cache has a
shared copy of the data in the next level cache.
[0151] Example 20 includes the apparatus of claim 14, wherein the
next level cache comprises a non-inclusive last level cache.
[0152] Example 21 includes a cache controller apparatus, comprising
means for identifying data from a working set for dynamic inclusion
in a next level cache based on an amount of re-use of the next
level cache, means for sending a shared copy of the identified data
to a requesting core of one or more processor cores, and means for
maintaining a copy of the identified data in the next level
cache.
[0153] Example 22 includes the apparatus of claim 21, further
comprising means for determining dynamic inclusion of data in the
next level cache on a per data line basis.
[0154] Example 23 includes the apparatus of claim 21, further
comprising means for incrementing a counter value when a hit in the
next level cache corresponds to an eviction from a core cache, and
means for identifying a current data hit in the next level cache
for dynamic inclusion in the next level cache if the current data
hit corresponds to an eviction from the core cache and if the
counter value is greater than a threshold.
[0155] Example 24 includes the apparatus of claim 23, further
comprising means for setting a snoop filter to indicate that the
requesting core is valid for the current data hit.
[0156] Example 25 includes the apparatus of claim 24, wherein, if
the current data hit does not correspond to an eviction from the
core cache or if the counter value is not greater than the
threshold, the method further comprises means for sending an
exclusive copy of the data to the requesting core, means for
updating an entry in the snoop filter to indicate a core identifier
of the requesting core, and means for deallocating the data in the
next level cache.
[0157] Example 26 includes the apparatus of claim 21, further
comprising means for silently dropping data to be evicted from a
core cache if the data to be evicted from the core cache has a
shared copy of the data in the next level cache.
[0158] Example 27 includes at least one non-transitory machine
readable medium comprising a plurality of instructions that, in
response to being executed on a computing device, cause the
computing device to identify data from a working set for dynamic
inclusion in a next level cache based on an amount of re-use of the
next level cache, send a shared copy of the identified data to a
requesting core of one or more processor cores, and maintain a copy
of the identified data in the next level cache.
[0159] Example 28 includes the at least one non-transitory machine
readable medium of claim 27, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to determine dynamic inclusion
of data in the next level cache on a per data line basis.
[0160] Example 29 includes the at least one non-transitory machine
readable medium of claim 27, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to increment a counter value
when a hit in the next level cache corresponds to an eviction from
a core cache, and identify a current data hit in the next level
cache for dynamic inclusion in the next level cache if the current
data hit corresponds to an eviction from the core cache and if the
counter value is greater than a threshold.
[0161] Example 30 includes the at least one non-transitory machine
readable medium of claim 29, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to set a snoop filter to
indicate that the requesting core is valid for the current data
hit.
[0162] Example 31 includes the at least one non-transitory machine
readable medium of claim 30, comprising a plurality of further
instructions that, in response to being executed on the computing
device, if the current data hit does not correspond to an eviction
from the core cache or if the counter value is not greater than the
threshold, cause the computing device to send an exclusive copy of
the data to the requesting core, update an entry in the snoop
filter to indicate a core identifier of the requesting core, and
deallocate the data in the next level cache.
[0163] Example 32 includes the at least one non-transitory machine
readable medium of claim 27, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to silently drop data to be
evicted from a core cache if the data to be evicted from the core
cache has a shared copy of the data in the next level cache.
[0164] Example 33 includes an integrated circuit, comprising a
core, and a cache controller coupled to the core, the cache
controller including circuitry to identify single re-use data
evicted from a core cache, and retain the identified single re-use
data in a next level cache based on an overall re-use of the next
level cache.
[0165] Example 34 includes the integrated circuit of claim 33,
wherein a source of the single re-use data is main memory.
[0166] Example 35 includes the integrated circuit of claim 34,
wherein the circuitry is further to determine a use count for a
data line based on a number of core cache hits experienced by the
data line when the data line is resident in the core cache,
determine a trip count for the data line based on a number of trips
made by the data line between the core cache and the next level
cache from when the data line is brought into one or more of the
core cache and the next level cache until the data line is evicted
from the next level cache, and identify the single re-use data
based on a use count of one and trip count of zero.
[0167] Example 36 includes the integrated circuit of claim 33,
wherein the circuitry is further to increment a counter value when
a hit in the next level cache corresponds to an eviction from the
core cache.
[0168] Example 37 includes the integrated circuit of claim 36,
wherein the circuitry is further to evict a data line from the core
cache, mark the evicted data line as dead, and install the evicted
data line marked as dead as a most recently used data line in the
next level cache if the counter value is greater than a threshold
and if a source of the data line is main memory.
[0169] Example 38 includes the integrated circuit of claim 37,
wherein, if the counter value is not greater than the threshold or
if a source of the data line is not main memory, the circuitry is
further to install the evicted data line marked as dead as a least
recently used data line in the next level cache if an invalid block
is available in the next level cache.
[0170] Example 39 includes the integrated circuit of claim 37,
wherein, if the counter value is not greater than the threshold or
if a source of the data line is not main memory, the circuitry is
further to bypass the next level cache if an invalid block is not
available in the next level cache.
[0171] Example 40 includes a method of controlling a cache,
comprising identifying single re-use data evicted from a core
cache, and retaining the identified single re-use data in a next
level cache based on an overall re-use of the next level cache.
[0172] Example 41 includes the method of claim 40, wherein a source
of the single re-use data is main memory.
[0173] Example 42 includes the method of claim 41, further
comprising determining a use count for a data line based on a
number of core cache hits experienced by the data line when the
data line is resident in the core cache, determining a trip count
for the data line based on a number of trips made by the data line
between the core cache and the next level cache from when the data
line is brought into one or more of the core cache and the next
level cache until the data line is evicted from the next level
cache, and identifying the data line as single re-use data based on
a use count of one and trip count of zero.
[0174] Example 43 includes the method of claim 40, further
comprising incrementing a counter value when a hit in the next
level cache corresponds to an eviction from the core cache.
[0175] Example 44 includes the method of claim 43, further
comprising evicting a data line from the core cache, marking the
evicted data line as dead, and installing the evicted data line
marked as dead as a most recently used data line in the next level
cache if the counter value is greater than a threshold and if a
source of the data line is main memory.
[0176] Example 45 includes the method of claim 44, wherein, if the
counter value is not greater than the threshold or if a source of
the data line is not main memory, the method further comprises
installing the evicted data line marked as dead as a least recently
used data line in the next level cache if an invalid block is
available in the next level cache, and bypassing the next level
cache if an invalid block is not available in the next level
cache.
[0177] Example 46 includes an apparatus, comprising one or more
processor cores, a core cache co-located with and communicatively
coupled to the one or more processor cores, a next level cache
co-located with and communicatively coupled to the core cache and
the one or more processor cores, and a cache controller co-located
with and communicatively coupled to the core cache, the next level
cache, and the one or more processor cores, the cache controller
including circuitry to identify single re-use data evicted from the
core cache, and retain the identified single re-use data in the
next level cache based on an overall re-use of the next level
cache.
[0178] Example 47 includes the apparatus of claim 46, wherein a
source of the single re-use data is main memory.
[0179] Example 48 includes the apparatus of claim 47, wherein the
circuitry is further to determine a use count for a data line based
on a number of core cache hits experienced by the data line when
the data line is resident in the core cache, determine a trip count
for the data line based on a number of trips made by the data line
between the core cache and the next level cache from when the data
line is brought into one or more of the core cache and the next
level cache until the data line is evicted from the next level
cache, and identify the single re-use data based on a use count of
one and trip count of zero.
[0180] Example 49 includes the apparatus of claim 46, wherein the
circuitry is further to increment a counter value when a hit in the
next level cache corresponds to an eviction from the core
cache.
[0181] Example 50 includes the apparatus of claim 49, wherein the
circuitry is further to evict a data line from the core cache, mark
the evicted data line as dead, and install the evicted data line
marked as dead as a most recently used data line in the next level
cache if the counter value is greater than a threshold and if a
source of the data line is main memory.
[0182] Example 51 includes the apparatus of claim 50, wherein, if
the counter value is not greater than the threshold or if a source
of the data line is not main memory, the circuitry is further to
install the evicted data line marked as dead as a least recently
used data line in the next level cache if an invalid block is
available in the next level cache.
[0183] Example 52 includes the apparatus of claim 50, wherein, if
the counter value is not greater than the threshold or if a source
of the data line is not main memory, the circuitry is further to
bypass the next level cache if an invalid block is not available in
the next level cache.
[0184] Example 53 includes a cache controller apparatus, comprising
means for identifying single re-use data evicted from a core cache,
and means for retaining the identified single re-use data in a next
level cache based on an overall re-use of the next level cache.
[0185] Example 54 includes the apparatus of claim 53, wherein a
source of the single re-use data is main memory.
[0186] Example 55 includes the apparatus of claim 54, further
comprising means for determining a use count for a data line based
on a number of core cache hits experienced by the data line when
the data line is resident in the core cache, means for determining
a trip count for the data line based on a number of trips made by
the data line between the core cache and the next level cache from
when the data line is brought into one or more of the core cache
and the next level cache until the data line is evicted from the
next level cache, and means for identifying the data line as single
re-use data based on a use count of one and trip count of zero.
[0187] Example 56 includes the apparatus of claim 53, further
comprising means for incrementing a counter value when a hit in the
next level cache corresponds to an eviction from the core
cache.
[0188] Example 57 includes the apparatus of claim 56, further
comprising means for evicting a data line from the core cache,
means for marking the evicted data line as dead, and means for
installing the evicted data line marked as dead as a most recently
used data line in the next level cache if the counter value is
greater than a threshold and if a source of the data line is main
memory.
[0189] Example 58 includes the apparatus of claim 57, wherein, if
the counter value is not greater than the threshold or if a source
of the data line is not main memory, the circuitry is further to
means for installing the evicted data line marked as dead as a
least recently used data line in the next level cache if an invalid
block is available in the next level cache, and means for bypassing
the next level cache if an invalid block is not available in the
next level cache.
[0190] Example 59 includes at least one non-transitory machine
readable medium comprising a plurality of instructions that, in
response to being executed on a computing device, cause the
computing device to identify single re-use data evicted from a core
cache, and retain the identified single re-use data in a next level
cache based on an overall re-use of the next level cache.
[0191] Example 60 includes the at least one non-transitory machine
readable medium of claim 59, wherein a source of the single re-use
data is main memory.
[0192] Example 61 includes the at least one non-transitory machine
readable medium of claim 60, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to determine a use count for a
data line based on a number of core cache hits experienced by the
data line when the data line is resident in the core cache,
determine a trip count for the data line based on a number of trips
made by the data line between the core cache and the next level
cache from when the data line is brought into one or more of the
core cache and the next level cache until the data line is evicted
from the next level cache, and identify the data line as single
re-use data based on a use count of one and trip count of zero.
[0193] Example 62 includes the at least one non-transitory machine
readable medium of claim 59, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to increment a counter value
when a hit in the next level cache corresponds to an eviction from
the core cache.
[0194] Example 64 includes the at least one non-transitory machine
readable medium of claim 63, comprising a plurality of further
instructions that, in response to being executed on the computing
device, cause the computing device to evict a data line from the
core cache, mark the evicted data line as dead, and install the
evicted data line marked as dead as a most recently used data line
in the next level cache if the counter value is greater than a
threshold and if a source of the data line is main memory.
[0195] Example 65 includes the at least one non-transitory machine
readable medium of claim 64, comprising a plurality of further
instructions that, in response to being executed on the computing
device, if the counter value is not greater than the threshold or
if a source of the data line is not main memory, cause the
computing device to install the evicted data line marked as dead as
a least recently used data line in the next level cache if an
invalid block is available in the next level cache, and bypass the
next level cache if an invalid block is not available in the next
level cache.
[0196] Reference in the specification to "one embodiment" or "an
embodiment" means that a particular feature, structure, or
characteristic described in connection with the embodiment is
included in at least one embodiment of the invention. The
appearances of the phrase "in one embodiment" in various places in
the specification are not necessarily all referring to the same
embodiment.
[0197] Some portions of the detailed description herein are
presented in terms of algorithms and symbolic representations of
operations on data bits within a computer memory. These algorithmic
descriptions and representations are the means used by those
skilled in the computing arts to most effectively convey the
substance of their work to others skilled in the art. An algorithm
is here, and generally, conceived to be a self-consistent sequence
of steps leading to a desired result. The steps are those requiring
physical manipulations of physical quantities. Usually, though not
necessarily, these quantities take the form of electrical or
magnetic signals capable of being stored, transferred, combined,
compared, and otherwise manipulated. It has proven convenient at
times, principally for reasons of common usage, to refer to these
signals as bits, values, elements, symbols, characters, terms,
numbers, or the like.
[0198] It should be borne in mind, however, that all of these and
similar terms are to be associated with the appropriate physical
quantities and are merely convenient labels applied to these
quantities. Unless specifically stated otherwise as apparent from
the discussion herein, it is appreciated that throughout the
description, discussions utilizing terms such as "processing" or
"computing" or "calculating" or "determining" or "displaying" or
the like, refer to the action and processes of a computer system,
or similar electronic computing device, that manipulates and
transforms data represented as physical (electronic) quantities
within the computer system's registers and memories into other data
similarly represented as physical quantities within the computer
system memories or registers or other such information storage,
transmission or display devices.
[0199] Certain embodiments also relate to apparatus for performing
the operations herein. This apparatus may be specially constructed
for the required purposes, or it may comprise a general purpose
computer selectively activated or reconfigured by a computer
program stored in the computer. Such a computer program may be
stored in a computer readable storage medium, such as, but is not
limited to, any type of disk including floppy disks, optical disks,
CD-ROMs, and magnetic-optical disks, read-only memories (ROMs),
random access memories (RAMs) such as dynamic RAM (DRAM), EPROMs,
EEPROMs, magnetic or optical cards, or any type of media suitable
for storing electronic instructions, and coupled to a computer
system bus.
[0200] The algorithms and displays presented herein are not
inherently related to any particular computer or other apparatus.
Various general purpose systems may be used with programs in
accordance with the teachings herein, or it may prove convenient to
construct more specialized apparatus to perform the required method
steps. The required structure for a variety of these systems will
appear from the description herein. In addition, certain
embodiments are not described with reference to any particular
programming language. It will be appreciated that a variety of
programming languages may be used to implement the teachings of
such embodiments as described herein.
[0201] Besides what is described herein, various modifications may
be made to the disclosed embodiments and implementations thereof
without departing from their scope. Therefore, the illustrations
and examples herein should be construed in an illustrative, and not
a restrictive sense. The scope of the invention should be measured
solely by reference to the claims that follow.
* * * * *